In alignment with the objectives of the S4H project, the collaborative efforts of the S4H team, in partnership with Dr. Amir Banimahd recently published a paper in the Journal of Earthquake Engineering and Structural Dynamics. This publication intricately delves into the team’s investigation into the efficacy of advanced machine learning (ML) techniques, specifically artificial neural networks (ANN), in predicting rocking block response. 

The research employs rigid blocks of varying sizes and slenderness to undergo pure rocking motion during seismic events, ensuring they neither slide nor bounce. The blocks either overturn or return safely to their original position after the ground shaking terminates. An artificial neural network (ANN) is trained to classify this response efficiently as either overturning or safe rocking based on structural parameters, ground motion characteristics, and the coefficient of restitution. The findings highlight the significant influence of ground motion velocity and frequency on overturning. Furthermore, the ANN is utilised to predict response amplitude and identify key input variables governing safe rocking, revealing the combined influence of ground motion duration, frequency, and intensity characteristics on rocking amplitude. Notably, the maximum incremental velocity (MIV), a novel intensity measure in rocking literature, demonstrates a significant correlation with rocking amplitude. The paper proposes closed-form expressions utilising the most influential input variables for quick yet sufficiently accurate response prediction. Lastly, emphasis is placed on the contribution of the coefficient of restitution, indicating its lesser impact on peak safe rocking response but heightened importance for overturning response.

The published version of the paper can also be accessed through Link.